This invention relates to a cross-connect switch for fiber-optic communication networks. In particular, this invention relates to a compact, multi-channel free-space optical cross-connect whereby switching is accomplished by tilting pairs of dual-axis micromirrors. Emergent communications systems use optical transmission through silica fibers using wavelength-division multiplexed (WDM) networks. As these systems evolve, requirements for reduced cost, form-factor, and power dissipation along with increased performance, scalability, and reliability become important in the design of efficient optical cross-connect systems. One particular type of optical cross-connect system utilizing dual-axis tilting micromirrors has been increasingly regarded as a technology which provides a solution to these constraints.
An example of prior art is Hoen, U.S. Pat. No. 6,253,001, which describes a free-space optical switch in which a plurality of collimated parallel beams is directed from a first two-dimensional fiber and lens array onto a first two-dimensional array of mirrors. Referring to FIG. 1, an unfolded optical switch 10 is shown as in Hoen having a collimator set 12, a first mirror set 14, a second mirror set 16, a second collimator set 18. In a specific configuration, sets 12 and 14 are configured to be inputs and sets 16 and 18 are configured to be outputs. A non-blocking optical switch is formed by directing collimated light from input set 12 onto the first dual-axis micromirror set 14. Light from the first micromirror set 14 is redirected to the second micromirror set 16, which redirects light into output collimator set 18. For optimal coupling, each individual collimator is assigned to a corresponding individual micromirror, for instance collimators 20, 22, and 24 are assigned to micromirrors 30, 32, and 34. To achieve an arbitrary non-blocking switch, an input micromirror, such as 30, must be able to swivel through a full range of motion to access all output mirrors, such as 32 and 34.
By tilting any two mirrors on the input and output mirror arrays, a non-blocking N×N cross-connect switch is established. For an N-port system, 2N fibers, lenses, and mirrors are required.
The fabrication and actuation methods of the micromirror arrays are key drivers for system cost and complexity. The fabrication methods for the mirror arrays involve either assembling discrete components or creating the arrays in parallel using batch fabrication techniques. Assembly of discrete components is an option for lower port-count switches, but it is generally not considered to he an appropriate cost-effective fabrication technique for larger port-count switches. For larger port-count switches, batch processing using advanced microfabrication techniques is an attractive alternative. These devices are referred to as Micro-Electrical Mechanical Systems (MEMS). The actuation methods of the MEMS mirrors typically fall into two categories: electrostatic and electromagnetic. Electromagnetic operation is generally used for large, discrete mirrors, because of the large forces that can be obtained. However, electromagnetic forces do not scale well for micro-devices. Electromagnetic actuation is challenging due to cross talk resulting from the difficulty of confining magnetic fields. In addition, high continuous currents, hysteresis, and immature processing techniques of magnetic materials call into question the reliability of electromagnetic operation. These constraints make it difficult to engineer compact, low-power electromagnetically actuated mirror arrays.
Electrostatic forces scale well for micro-devices. Electrostatic actuation techniques fall into two major categories, comb-drive actuation and parallel-plate actuation. In the case of comb-drive actuation, comb-drive actuators develop forces between interdigitated combs that are located away from the mirror by the use of linkage elements, which arc typically in contact with each other. Although this technique has the advantage of decoupling electrostatic forces from the mirror design, allowing conceivably lower voltages for a given force, there are several significant disadvantages. These disadvantages include issues of compactness, difficulty in manufacturing, difficulty in interconnection, and the potential for undesired contact with adjacent components and regions.
Parallel-plate actuation overcomes many of the limitations of the other actuation methods. This actuation method utilizes non-contacting structures where the electrostatic forces are developed between the mirror and the lower electrodes. This actuation method avoids the reliability issues associated with contact. With backside interconnects, it can be engineered to be compact. Because the actuation is electrostatic, it is also low-power. However this technique typically requires higher voltages.
With all actuation techniques, there are trade-offs between tilt angle, switching time, voltage, and optical efficiency which make it desirable to limit the maximum tilt angle of the mirrors. For optical network restoration reasons, optical switches are generally required to have switching times on the order of 10 ms. Typical 256 port free-space optical designs require the mirror tilt angle reproducibility to be on the order of 1 part in 10,000. Assuming mirrors with highly damped fundamental torsional modes of oscillation, a minimum switching time of approximately one oscillation period is theoretically possible. This results in mirrors with a fundamental torsional resonant mode of about 100 Hz. It is difficult to maintain the stability required over the lifetime of a product. Even with a fully closed loop monitoring system, the calibration of the monitoring system can easily vary more than one part in 10,000.
As a result, optical monitoring of the coupled output power is generally required. A combination open-loop/closed-loop feedback system consisting of a single optical output power tap per output mirror may be used. In this system mirrors are steered open-loop to a position where power is coupled into the output fiber, although it may not be optimally coupled. Once light has been coupled into the output fiber, the mirrors are positioned using a servo control system to maximize the efficiency of the coupled light. In this case, a much more reasonable mirror reproducibility of approximately 1 part in 100 is required, which has been demonstrated to fall well within the capabilities of electrostatically actuated MEMS devices. With closed loop operation, a minimum of 3 to 10 cycles are generally required to capture and servo to maximum power. Because of this, mirrors with a fundamental resonant mode of 300 Hz to 1 kHz are generally required to achieve 10 ms switching times. Such mirrors have stiff hinges, which require significant electrostatic force to achieve a given tilt angle.
The maximum tilt angle of a parallel-plate actuated MEMS device is closely coupled with the maximum voltage that may be applied between the electrodes. Trade-offs between tilt angle, resonant frequency, voltage, and optical efficiency make it desirable to minimize the tilt angles of the mirrors.
Inventions are known wherein oval mirrors are disposed in arrays in a three dimensional switch similar to that shown in FIG. 1. Reference is made to PCT publication WO 01/92942 dated 6 Dec. 2001 in the name of Applicant Calient Networks. Therein the oval mirrors arc provided to accommodate beamlets which project with oval footprints. Therein the axis of each micromirror that is perpendicular to the plane of an incident and reflected beamlet is parallel to the minor axis of the micromirror. This is done to permit greater rotation of the micromirror around the major axis of the micromirror. As a consequence, the arrays are designed to fit within an oval boundary that has it major axis transverse and generally perpendicular to the plane of the incident and reflected beamlets, that is, it was wider than its length. As a result, it has been discovered that micromirrors near one periphery must be rotatably displaced much greater than others near an opposing periphery to address all micromirror locations. There is no suggestions of minimization of tilt angles.
Byway of background, the concept of optical MEMS switches with converging beams are known. Examples of how to realize converging beamlets is taught in Published U.S. patent application Publication No. US 2003/0002783 published Jan. 2, 2003 in the name of Lucent Technologies, Inc. of Holmdel, N.J.
What is needed is a solution which globally minimizes the maximum tilt angles of all the micromirrors while still maintaining a compact, efficient, and cost-effective optical switch.